Every decision delayed for the last 30 percent of information pays for that information in time, and time is the one input an operator cannot buy back. That is the cost of certainty, and most organizations pay it invisibly, every week, on every decision that sits in a queue waiting for one more analysis, one more data pull, one more meeting to confirm what the last meeting already showed. Nobody books the expense. It books itself: in options that expire while the deck gets refined, in competitors who moved while the committee reconvened, in teams that learn to wait because waiting is what leadership models.
Decision velocity is not a personality trait. It is an operating tempo, it is engineerable, and the 70 percent rule for decision making is the complete framework for engineering it.
- The Rule, Stated Precisely
- The Decision Classification Matrix: Not All Decisions Get 70
- The Three-Question Test: Knowing When You Are at 70
- The Time-Value Calculation: Why the Math Favors 70
- Running the Rule in an Organization
- The Rule Inside the Transformation
- Frequently Asked Questions
- About the Stagnation Assassin
The Rule, Stated Precisely
The 70% Rule says to act on most business decisions at roughly 70 percent of ideal information. Ideal information is what unlimited time and budget could gather. The threshold is calibrated confidence verified by a repeatable test, not felt certainty, and it flexes up and down by decision class.
Every word in that statement is load-bearing, so define them. Ideal information is what a decision-maker with unlimited time and budget could gather, not omniscience; 70 percent of ideal is a high bar that most delayed decisions passed weeks before anyone acted on them. Roughly matters because 70 is a central tendency, not a magic number, and the framework below calibrates it up and down by decision class. And confidence here means calibrated confidence, not feeling: the rule is about having 70 percent of the obtainable information, verified by a test you can run, not about waiting until your stomach reports 70 percent sure. Feelings are not calibration. The distinction between calibrated confidence and felt certainty is the intellectual spine of the whole framework, and it is what separates this rule from the pop version that circulates as a permission slip for shooting from the hip.
Why 70 and not 50 or 90? Because 70 percent is the approximate point at which, on most business decisions, the marginal information value of more analysis falls below the marginal opportunity cost of more waiting. The information value of additional analysis is real, but it is bounded, and on most decisions it runs out well before confidence reaches 95. Past that point the slope inverts: more analysis produces diminishing information while opportunity cost accumulates without limit. The operator who recognizes the inversion stops analyzing where information value runs out. The operator who does not keeps analyzing until the conditions that made the decision relevant have changed, at which point the decision has been overtaken by events and the operation has paid the full cost of the delay for nothing.
I learned the rule the way I learned everything in this methodology: on the clock, with the meter running. The morning after the Tuesday night that produced the Growth Trap diagnosis, I had a hypothesis package of five structural moves, each with its own internal logic, none validated to 95 percent confidence on any single dimension. What I had was roughly 70 percent confidence on the package, and that was enough to act, because the alternative was six more months of the cost cascade compressing operating margin and six more months of organizational drift making the eventual execution harder.
The choice was never between acting at 70 percent and acting at 95 percent. It was between acting at 70 percent now and acting at 95 percent never, because the alternative was six more months of losing roughly a million dollars a quarter.
By the time 95 arrived, the conditions that produced the diagnostic would have changed, and the operation would have lost the financial capacity to fund the moves it required.
The Decision Classification Matrix: Not All Decisions Get 70
The Decision Classification Matrix scores every decision on two dimensions, reversibility and criticality, producing four classes. Class A acts at 50 to 60 percent confidence, Class B at 70, Class C at 70 with engineered exits, and Class D at 85 to 90 percent.
The rule’s most important refinement is that 70 percent is the default, not the universal. Decisions vary across two dimensions that together set the right threshold.
Dimension one: reversibility. A reversible decision can be undone at limited cost if it proves wrong: a pricing change that rolls back, a vendor that gets replaced, a staffing assignment that gets reassigned. Reversibility absorbs the cost of error, because the error gets corrected before it compounds, so the threshold can sit lower. An irreversible decision cannot be undone, or only at substantial cost: a major acquisition, a facility closure, a senior leadership change, a capital deployment that locks a multi-year pattern.
Dimension two: criticality. A critical decision affects a material portion of the operation’s structural slope. A non-critical decision affects a bounded slice: the routine vendor selection, the standard hire, the minor process change, where even a wrong answer carries a capped downside.
| Class | Reversibility | Criticality | Confidence Threshold | Operating Discipline |
|---|---|---|---|---|
| Class A | Reversible | Non-critical | 50 to 60% | Decide within 48 hours |
| Class B | Reversible | Critical | 70% | Three-Question Test, document reversibility |
| Class C | Irreversible | Non-critical | 70% | Engineer contractual or operational exits |
| Class D | Irreversible | Critical | 85 to 90% | Pilot, outside perspective, sensitivity testing |
Class A: Reversible and Non-Critical
The bulk of an operator’s week: standard pricing within established bands, routine vendor selections, operational adjustments inside existing processes. Threshold: 50 to 60 percent. Reversibility absorbs the cost of error and non-criticality bounds it, so these decisions get made inside forty-eight hours, and the rare miss gets corrected when the data comes in. An organization that runs Class A decisions through Class D process has built a queue where a tempo should be.
Class B: Reversible and Critical
Decisions that affect a material part of the operation but can be undone if wrong: pricing changes on major relationships, significant portfolio changes, sales-territory restructuring. This is where the 70 percent threshold lives. The stakes justify more confidence than Class A; the reversibility absorbs the residual risk; and the operation captures the opportunity value of moving at tempo. Most of the decisions that determine a transformation’s speed are Class B, which is why 70 is the framework’s central tendency.
Class C: Irreversible and Non-Critical
Cannot be undone, but bounded in scope: the long-term vendor contract on a non-core service, the platform selection for one functional area, the facility lease for a non-critical operation. Threshold: 70 percent, with engineered exits. The irreversibility argues for more care than Class B, but the bounded criticality means the answer is not more analysis. It is structural protection: move at 70 and build contractual or operational exit clauses that cap the cost of being wrong.
Class D: Irreversible and Critical
The decisions that lock in the structural slope permanently: major acquisitions, facility closures, senior leadership changes, multi-year capital deployments. Threshold: 85 to 90 percent. Both dimensions compound the cost of error, so this is the one place the operator does not default to 70. Invest in the additional analysis, run the pilot, bring in the outside perspective, and reach the higher confidence before executing.
One more discipline: classifying the decision is itself part of the work, because the class is not always obvious in the moment. A decision that looks reversible may carry hidden lock-in. A decision that looks non-critical may sit on a slope the operator has not fully recognized. Run the classification explicitly, out loud, before setting the threshold. Intuition about which threshold applies is how Class D decisions get made at Class A speed, and how Class A decisions get Class D committees.
The Three-Question Test: Knowing When You Are at 70
The Three-Question Test measures whether you have reached 70 percent confidence. Can you name the material risks? Can you explain the decision to an intelligent outsider? Can you state a testable hypothesis about the outcome? Three yes answers means the threshold is met: act.
The rule’s practical problem is measurement: confidence does not come with a gauge. The Three-Question Test is the gauge, and passing it means the threshold is reached.
Question One: Do You Understand the Key Risks and Downsides?
Not every risk. The material ones, the two or three that could meaningfully impair the decision’s value if they materialize. An operator who can articulate them has done the analysis that produces the upper bound of confidence. An operator who cannot is working at lower confidence than they realize, however much data is on the desk.
Question Two: Can You Explain the Decision Clearly to Someone Outside the Situation?
Not the specialist who already knows the context. An intelligent outsider, not steeped in the detail. A clean explanation means the decision has been internalized well enough to act on. A muddy one means the analysis is incomplete, regardless of its volume.
Question Three: Do You Have a Reasonable Hypothesis About What Will Happen?
Not a certainty. A hypothesis specific enough to be tested against the actual outcome once the decision executes. An operator who has one has converted analysis into a directional prediction, which is the threshold for action. An operator who cannot state one is still in the analysis phase, whatever the data count says.
Three questions, three yes answers, act. And here is the discipline that makes the test a rule instead of a ritual: the test is the threshold, not a checkpoint on the way to a higher one. Pass it on a Class B decision and you execute. Keep gathering analysis after passing it and you are working against the rule, accumulating opportunity cost without producing meaningful confidence. The one exception is Class D, where the test is necessary but not sufficient: pass it, then validate the hypothesis through pilots, outside perspectives, and sensitivity testing on the key assumptions, because there the compounded cost of error is what the higher threshold protects against.
The Time-Value Calculation: Why the Math Favors 70
The time-value calculation compares three terms: expected value at 95 percent confidence, expected value at 70 percent, and the cost of the delay between them. On most business decisions the delay cost exceeds the value gap, so acting at 70 captures more total value.
The rule is mathematical, not aesthetic, and the calculation has three terms. Term one: the expected value of the decision made at 95 percent confidence, which is, on average, slightly higher than at 70, because the extra analysis trimmed residual uncertainty. Term two: the expected value at 70 percent, slightly lower, because the probability of being wrong is correspondingly higher. Term three: the cost of the delay between the two points, which includes the value of the decision itself for the weeks or months it sits unmade, the competitive position that shifts during the wait, the team’s attention consumed by analysis instead of execution, and the organizational momentum that erodes when decisions queue instead of resolve.
The rule says that on most decisions, term three exceeds the gap between terms one and two. Act at 70 and you capture more total value than waiting for 95, even after adjusting for the higher probability of being wrong at the lower threshold.
And the arithmetic is class-dependent, which is why the matrix exists. On reversible decisions, the cost of being wrong is capped by the cost of reversing, usually low, so the 70 percent action captures the upside while reversibility absorbs the downside, and the analysis that would push to 95 produces marginal information dwarfed by the time it costs. On irreversible decisions the arithmetic shifts: the cost of error is uncapped by reversibility, the higher error probability at 70 produces a larger expected loss, and the additional analysis earns its keep. Same formula, different inputs, different thresholds. The framework is one calculation applied honestly, not one number applied everywhere.
The cart company resolution arc was this calculation, run repeatedly, in public. The mix shift to remanufactured carts, the pricing audit on the manual-line cascade, the plant conversion: each a Class B decision, reversible at moderate cost, critical to the slope, made at roughly 70 percent, executed within weeks. The bigger cart was the exception that proves the calibration: a Class C decision with significant capital and a long development commitment, so the confidence bar moved. End-user research with the secondary-banner operators, prototype testing, financial modeling on the fleet-replacement dynamic, cumulative analysis pushing confidence to roughly 80 percent before authorization. Two different thresholds for two different classes, both from one framework. That is what “structured” means.
Running the Rule in an Organization
Deploy the rule by decision class. Delegate Class A decisions with a 48-hour expectation, delegate Class B with the Three-Question Test attached, review Class C exit clauses, and publish the Class D list. Review tempo and hypotheses, never re-litigate individual calls, and budget errors openly.
Here is the part the pop versions of this rule skip entirely, and it is the part that determines whether the rule changes anything: a 70 percent rule that lives only in the operator’s head caps the organization’s decision velocity at the operator’s calendar. The framework’s real payoff comes from deployment, pushing the rule down through decision rights so the whole organization moves at engineered tempo. This is the operational substance behind the case that even Harvard Business Review now makes for making great decisions quickly: speed is a system property, not a leadership mood.
Deploy by class. The delegation architecture writes itself off the matrix: Class A decisions get delegated wholesale, with the 48-hour expectation attached. Class B decisions get delegated to the leaders who own the affected slope, with the Three-Question Test as the required discipline and the reversibility documented as part of the decision. Class C requires the exit-clause engineering, which is reviewable. Class D stays with the operator, and everybody knows which decisions those are in advance, because a Class D list published in daylight prevents both the hoarding of routine decisions and the quiet delegation of existential ones. If the constraint on your business is that every decision routes through one calendar, this architecture is the fix, and the condition has a name: the Founder Bottleneck.
Review velocity without re-litigating decisions. The operating review asks two questions about the decision system: are decisions being made at the tempo their class prescribes, and are the hypotheses from Question Three being tested against outcomes? It does not re-argue individual calls, because a review that re-litigates teaches everyone to route decisions back upward, and the queue rebuilds itself inside a quarter.
Budget the errors, out loud.
A team running Class B decisions at 70 percent confidence will be wrong at a knowable rate. That is not a defect in the system. That is the system, and the error rate is the price of the tempo, paid deliberately.
The time-value calculation says the tempo is worth more, and the corollary is non-negotiable: punishing a calibrated miss, a decision that passed the Three-Question Test, documented its hypothesis, and drew a bad card, destroys the tempo permanently, because the organization learns that the real threshold is 100 percent and starts padding every decision with protective analysis. Review calibrated misses for what the hypothesis got wrong. Reserve consequences for uncalibrated ones: decisions made without the test, without the hypothesis, without the classification. The difference between those two responses is the difference between an organization that decides and an organization that documents. Analysis paralysis, examined honestly, is usually an accountability problem wearing an information costume.
The Rule Inside the Transformation
Inside the transformation methodology, the 70% Rule sets the operating tempo of the 90-day playbook. Days 1 to 30 run reversible Class A and B moves, the pricing test executes as Class B, and the structural commitments of Days 61 to 90 carry the higher Class C and D thresholds.
In the transformation methodology, the 70% Rule is the operating tempo of the 90-day playbook: the unilateral leak-plugging moves of Days 1 to 30 go first precisely because they are Class A and B, reversible, clearing the bar without waiting; the pricing test runs as a Class B move with the elasticity data as its hypothesis check; and the structural commitments of Days 61 to 90 are the Class C and D calls that earned their higher thresholds. The rule is also the answer to the question every rebuilt operation eventually faces: the structural moves created the capacity to move fast, and decision velocity is what converts that capacity into captured value. A rebuilt machine run at committee tempo is a sports car in a school zone, permanently.
Where the rule gets misapplied, name it honestly: 70 percent is not a license for recklessness on the decisions that deserve 90, and it is not a personality cult of decisiveness. It is a calculation, run per class, tested per decision, deployed per decision right, and reviewed per hypothesis. If you have been quoting the famous shareholder-letter version of this idea for years, understand what that maxim shares with this system and what it lacks: the maxim names the threshold; the system supplies the classification, the test, the math, and the deployment that make the threshold operational.
Somewhere on your desk right now is a decision that passed the Three-Question Test two weeks ago and is still waiting, accruing opportunity cost daily, for a confidence level it will never reach and does not need. Classify it. Test it. If it passes: it was ready before you were. Act.
The 70% Rule is the decision-velocity framework of Ten Minute Transformation (Koehler Books, February 2027), where it pairs with the Automation Ceiling as the strategic-navigation layer that sets the rebuilt operation’s tempo.
Frequently Asked Questions
Executives ask three recurring questions about the 70% Rule: what it means, when not to use it, and how to know when enough information is in hand. The short answers: act at 70 percent of obtainable information, raise the bar to 85 to 90 percent on irreversible critical calls, and run the Three-Question Test.
What is the 70 percent rule in decision making?
The 70 percent rule says that on most business decisions, you should act at roughly 70 percent of ideal information, where ideal means everything a decision-maker with unlimited time and budget could gather. It replaces felt certainty with calibrated confidence, verified through the Three-Question Test and adjusted by decision class.
When should you not use the 70 percent rule?
On Class D decisions: irreversible and critical calls such as major acquisitions, facility closures, senior leadership changes, and multi-year capital deployments. Those decisions carry a threshold of 85 to 90 percent confidence, reached through pilots, outside perspectives, and sensitivity testing on the key assumptions.
How do you know when you have 70 percent of the information?
Run the Three-Question Test. Can you articulate the two or three material risks? Can you explain the decision clearly to an intelligent outsider? Can you state a testable hypothesis about what will happen? Three yes answers means the threshold is reached, and on reversible decisions, passing the test means you act.
About the Stagnation Assassin
Todd Hagopian is a Fortune 500 transformation executive who has generated $3B+ in shareholder value across Berkshire Hathaway, Illinois Tool Works, Whirlpool, and JBT Marel, where he serves as VP of Global Product Strategy. Known as The Stagnation Assassin, he is the author of two published books: The Unfair Advantage: Weaponizing the Hypomanic Toolbox and Stagnation Assassin: The Anti-Consultant Manifesto. His blog is published in 15+ languages and read by operators worldwide. Bring him to your stage via his speaking page or connect with him on LinkedIn.
How many decisions on your desk passed the Three-Question Test weeks ago and are still waiting? Book a 15-minute Decision Velocity Audit: we classify your ten biggest open decisions, assign each its threshold, and identify the ones that were ready before you were. Request the audit here. Bring your queue. Leave with a tempo.